How we audited a 40,000-line business platform, redesigned the online store and added 11 Indian languages, without changing a single rule in the live business.
Chavan Brothers grinds and sells masala from its shop in Lalbaug, Mumbai, and ships across India. Behind the shop is a custom Django platform that runs billing, custom grinding orders, barcode sales, purchasing, warehousing, delivery, payroll and the online store. JMDA Analytic was asked to understand that platform and give the online store a modern, multilingual front end, with no risk to day-to-day operations.
Chavan Brothers is one of Lalbaug's best-known masala shops. It began in 1937 as Kamgar Masala, beside the Digvijay Mill gate in Girangaon, and was named for the mill workers (kamgar) and office-goers whose kitchens it supplied.
The family has a habit of being first. Its taglines in Marathi are “स्वादिष्ट घरगुती मसाल्याचे माहेरघर” (the home of tasty homemade masala) and “विश्वास ग्राहकांचा आणि प्रदीर्घ अनुभव तब्बल तीन पिढ्यांचा” (customers' trust, and three generations of experience).
Late Vishnu Natha Chavan opens Kamgar Masala in Lalbaug, beside the Digvijay Mill gate and Western India Mill.
Brings electric iron pounders into Mumbai's masala trade, the first to do so, replacing foot-operated pounding. The shop becomes Chavan Brothers.
Present owner Shri Baban Vishnu Chavan becomes the first to remove chilli stalks before grinding, for a cleaner, tastier masala.
The third generation keeps the recipes and ships across Maharashtra, India and abroad, backed by its own software platform.
Our first job was to read the whole system and document it for the client. This is the module map we produced. Each box is a Django app; together they cover the business from purchase of raw spices to the customer's doorstep.
new_pagesHomepage, category and product pages, cart, address book, order history, policy pages and enquiry form.
online_productOnline cart, pack sizes, invoices, payment records and Worldline gateway responses.
userAccounts, OTP login by SMS and WhatsApp, addresses, states, pincodes, areas and warehouses.
invoiceMasala-type orders priced by pack size, labour charges, invoices, docket numbers, credit notes and sales reports.
barcode_pageBarcode labels, scan-to-cart billing, daily summary, purchase entries, loose and pack returns.
category_systemCategory-based carts and retail invoices for counter and multi-masala orders.
offline_moduleItem masters (raw, finished, third-party), vendors, customers, purchases, ledgers, expenses, warehouse moves, home production, couriers, monthly reports.
inventoryNightly stock snapshots for barcode packs, loose goods and shop stock, run by scheduled jobs.
dispatch_boyDelivery records and reports for dispatch staff, and payment links for collecting money at the door.
employee_appLabour documents, attendance, basic salary, increments, advances and payments.
pageEarlier website pages and the job that checks pending online payments every two minutes.
GSTNumberGST registration details used on invoices.
Customer picks a product and a pack size (100 g, 250 g, 1 kg…). Prices are MRP.
Mobile number only. A code arrives by SMS and WhatsApp. No password.
Saved billing and delivery address. The billing state sets the delivery rate.
Delivery is added by weight: ₹80 per kg in Maharashtra, ₹100 per kg elsewhere.
Secure Worldline payment page. A job re-checks pending payments every two minutes.
The order becomes an invoice and is finalised by the evening job.
Courier docket number, or local delivery by the shop's dispatch staff.
Courier follow-ups, order history for the customer, monthly sales reports for the owner.
The question we had to answer: how do you give an 89-year-old business a 2026 online store when you are not allowed to touch the engine?
Instead of editing the live homepage, we built new “test” pages that sit next to it at their own addresses. They read the same database and call the same cart, login and enquiry code as the live site. The owner can try them on real phones, and the live site never changes until the business decides to switch.
We traced every module, model and customer flow and wrote a codebase guide for the client: an app map, the order and payment flows, a diagram, and a prioritised list of 14 bugs and risks. Examples: duplicated copies of the same views in old folders, a delivery calculation that does not handle orders above 16 kg, and a fee row in the cart that can never apply.
We built a modern brand-led page (/test-home/) and a quick-commerce page (/test-quick/) in the style Indian shoppers know from Blinkit and similar apps. The owner chose quick-commerce. We then refined it in rounds: removed sections that didn't sell, checked every FAQ answer against how checkout actually works, and showed every product of every category with its price.
A language picker with English, Hindi, Marathi, Gujarati, Kannada, Tamil, Telugu, Bengali, Malayalam, Punjabi and Odia, plus Google Translate on demand for any other language.
We cut database work to a fixed handful of queries, sent product cards ready-made from the server and deferred off-screen sections. Each page also has a built-in setup check, so the shop's own team can install updates without breaking anything.
A new About page (/test-about/) telling the 1937 story in the same design and 11 languages, plus a plan for a Flutter app for Android and iPhone that reuses the same back end.
Everything a shopper needs is on the first screen: categories, bestsellers, price and pack size, and an ADD button. Every category then follows as its own row with every product in it.
ADD calls the site's existing add-to-cart API. Products with several sizes open a pack-size sheet. A sticky cart bar keeps the total in view.
If a shopper isn't logged in, ADD opens an OTP sheet on the same page, using the same login APIs as the live site.
Delivery rates, payment and returns copy was checked line by line against the real checkout code, and wrong claims were removed.
320, 360, 390, 414, 768, 820, 1024, 1280, 1440 and 1920 px, from small Android phones to large desktop monitors, in every language. That is 110 combinations, all passing with no overflow and no script errors.
Most Indian online stores are English only. For a brand whose customers think about food in Marathi, Hindi or Tamil, that is a barrier at the exact moment of buying.
The new About page tells the 1937 story as a timeline, explains the name “Kamgar Masala”, shows the shop's own product photos and ends with a clear path back to the store. It is in the same design and the same 11 languages.
Key dates (1937, 1965, 1990) up front, because heritage is the brand's strongest selling point.
Address, hours, map directions, both phone numbers, WhatsApp and email in one block.
Facts come from the existing site in English and Marathi. We flagged a wrong gallery caption for the client to correct.
We tested every customer-facing claim against the source code. Two examples: delivery is charged per started kilogram based on the billing state, and the payment-method fee described on the old FAQ never actually applies at checkout. The new pages say only what the system does.
| Technique | Effect |
|---|---|
| Fetch only needed columns | Smaller queries for the product list |
| Two bulk queries for all pack sizes and cart flags | No per-product database calls |
| Cards filled on the server | Prices visible before any script runs |
| Off-screen sections deferred | Browser lays out only what is near the screen |
| Fonts per language, on demand | No extra download for English visitors |
Measured on a test server with sample data. Live figures depend on the real catalogue and hosting.
| Deliverable | Value to Chavan Brothers |
|---|---|
| Codebase guide | A written map of all 12 modules, the order and payment flows, and 14 prioritised bugs and risks. Future developers can work safely from day one. |
Quick-commerce store (/test-quick/) | Every product and price on the homepage, one-tap add to cart, inline OTP login, sticky cart. Running on the live server and confirmed working by the client. |
| 11-language front end | Hand-translated store and About page, plus Google Translate for any other language. |
| Speed fixes | A fixed number of database queries regardless of catalogue size, and cards that appear before scripts load. |
About page (/test-about/) | The 1937 story as a sales tool, in the new design and all 11 languages. |
Alternative design (/test-home/) | A brand-led option kept for comparison and future A/B testing. |
| Install kit | Paste-in code, step-by-step install notes and server tips (for example the Apache locale fix for non-English text in error pages). |
| Mobile app plan | A Flutter plan for Android and iPhone that keeps checkout and payment on the proven web flow, plus a store-publishing checklist. |
The new pages are being piloted alongside the live site. We will measure these with the site's existing Google Analytics and order data, comparing the old and new pages:
Load time and PageSpeed score
Share of visits with an ADD tap
Visits that end in a paid order
Share of visitors who switch language
Visitors leaving from the first page
Online orders, call-backs, WhatsApp chats
Switch the homepage and About page to the new design once pilot numbers confirm the gain.
Strengthen login security and fix the logged issues, such as heavy-parcel delivery and leftover cart rows.
Serve resized product thumbnails so phones download a fraction of the image data.
Flutter app for Android and iPhone: native catalogue, the existing web checkout, push notifications.
A careful audit showed the existing platform was sound. A new front end on the same logic delivered the visible change at a fraction of the risk and cost.
Parallel test pages let the owner use the new store on real phones with real data before a single customer sees it.
For regional food brands, the customer's mother tongue is part of the product, not a translation afterthought.
Reading the checkout code caught FAQ promises the system didn't keep. Accurate copy builds trust at the moment of payment.
An 89-year-old Lalbaug masala house now has a fast, multilingual, phone-first store and a documented platform, built on the system it already trusted, with no change to how it bills, prices or gets paid.
Talk to JMDA Analytic about a no-risk modernisation, starting with an audit of what you already have.
Chavan Brothers names, history and photographs belong to Chavan Brothers. Screens show test builds with sample product data.